Up to date
Qwen Qwen License (qwen) out 2 years ago
72.7B parameters, text in, text out, 32K tokens of context. 2 providers serve it from $0.36 per million input tokens.
Verified 2026-10-26 07:27 UTC Rank #74 in LLMs · score 103.9 method v1 Sources 4 Downloads 30d 217.2k · #241 of 619 tracked
Changed this week: 4 new quantized builds on the Hub (98 in all).2026-10-24
Picked by rule from 32 tracked pieces: the lab's own post, the most shared, a field report, the latest.
The 2 picks above, plus these 23. Top is the radar's score: discussion, source authority and the lab's own posts, fading with age.
Live hosted endpoints, cheapest first, USD per million tokens. Ties break on output price, then provider name.
Hugging Face does not say who downloads a model or with what. These are the proxies: API providers listing it, quantized builds, fine-tunes/adapters/merges, and Spaces that load it.
Served by 2 API providers; heavily run locally (98 quantized builds); a big base for builders (176 fine-tunes, adapters and merges).
Top Spaces running it: gokaygokay/FLUX-Prompt-Generator ♥ 1,400 · allenai/reward-bench ♥ 438 · baconnier/prompt-plus-plus ♥ 246 · JournalistsonHF/ai-scraper ♥ 150
Weights only, an estimate: leave room for the KV cache. Method
About 44 GB at INT4, 87 GB at FP8, 175 GB at BF16: weights only, so leave room for the KV cache.
Estimate params × bytes per param × 1.2 · how we estimate
Weights, every dated article, and repo commits on one rail.
Teal is the weights (2024-09-16), amber triangles are 4 dated articles (each links to it), grey ticks are repo commits.
274 derivatives on the Hub. Which build to run, by runtime: the most downloaded in each class.
From the Hugging Face model tree; quantized builds in teal, fine-tunes, merges and adapters in greys.
Everything else, folded: the full tables, every source.
It supports a full context length of 131,072 tokens.[1]
Generated 2026-10-07 15:00 UTC by parasail-gemma-4-26b-a4b-it; each numeral points at the fact it rests on. Not a verdict.
Labels are evidence, not verdicts: measured a third-party leaderboard row for this model; author-reported a number the model author published in the card or eval files; structural a deterministic signal from the name, config or chat template, or a decision-model estimate. Each chip links to its source. All use cases.
Listed in LLMs
| Pipeline tag | text-generation |
|---|---|
| Library | transformers |
| Architectures | Qwen2ForCausalLM |
| Model type | qwen2 |
| Dtype | bfloat16 |
| Params by dtype | BF16: 72,706,203,648 |
| Size band | XL |
| MoE | no |
| Context length | 32,768 (config.json max_position_embeddings) |
| Languages | en |
| Papers | arXiv:2309.00071, arXiv:2407.10671 |
| License | Qwen License (qwen) (other) |
| Hub snapshot | 2026-10-26 |
vLLM: supported · Qwen2ForCausalLM · 4b1c04e76fe9ac3ca8b7e541dd423e7d5f38dc60SGLang: supported · Qwen2ForCausalLM · 60d941293e3af27a12c6eb64e2c3e2e1a17f100f
vLLM registry (4b1c04e76fe9ac3ca8b7e541dd423e7d5f38dc60) · SGLang registry (60d941293e3af27a12c6eb64e2c3e2e1a17f100f) · Matched by architecture class name against each project's registry on main; 'supported' means the architecture is registered, not that this checkpoint was tested.
Weights-only memory for Qwen2.5-72B-Instruct: params × bytes/param × 1.2 overhead. KV cache and activations are extra. How this is computed.
| BF16 (2 bytes/param) | 175 GB · Multi-GPU node (<640 GB) |
|---|---|
| FP8 (1 byte/param) | 87 GB · Multi-GPU node (<640 GB) |
| INT4 (0.5 bytes/param) | 44 GB · Single high-end GPU (<80 GB) |
Weights only (params × bytes/param × 1.2 overhead); KV cache and activations extra.
Base model: Qwen/Qwen2.5-72B
quantized: 98finetune: 71merge: 16adapter: 89
| Variant | Relation | Quant | Params | Band | License | Author | Likes | Downloads |
|---|---|---|---|---|---|---|---|---|
| Qwen/Qwen2.5-72B-Instruct-AWQ | quantized | AWQ | 73.0B | XL | other | Qwen | 91 | 12.4M |
| huihui-ai/Qwen2.5-72B-Instruct-abliterated | finetune | — | 72.7B | XL | other | huihui-ai | 86 | 1.4M |
| bartowski/Qwen2.5-72B-Instruct-GGUF | quantized | GGUF | — | — | other | bartowski | 48 | 693.6k |
| Qwen/Qwen2.5-72B-Instruct-GPTQ-Int4 | quantized | GPTQ | 73.0B | XL | other | Qwen | 51 | 533.3k |
| mlx-community/Qwen2.5-72B-Instruct-4bit | quantized | -4bit | 72.7B | XL | other | mlx-community | 7 | 315k |
| mesolitica/Qwen2.5-72B-Instruct-FP8 | quantized | -FP8 | 72.7B | XL | — | mesolitica | 0 | 229.4k |
| Qwen/Qwen2.5-72B-Instruct-GPTQ-Int8 | quantized | GPTQ | 73.0B | XL | other | Qwen | 36 | 210.4k |
| unsloth/Qwen2.5-72B-Instruct-bnb-4bit | quantized | -bnb | 74.9B | XL | other | unsloth | 8 | 197.4k |
| OpenGVLab/InternVL2_5-78B | merge | — | 78.4B | XL | other | OpenGVLab | 193 | 104.4k |
| Qwen/Qwen2.5-72B-Instruct-GGUF | quantized | GGUF | — | — | other | Qwen | 57 | 102.9k |
| Nexusflow/Athene-V2-Chat | finetune | — | 72.7B | XL | other | Nexusflow | 292 | 70k |
| zetasepic/Qwen2.5-72B-Instruct-abliterated-GGUF | quantized | GGUF | — | — | other | zetasepic | 7 | 57.1k |
| second-state/Qwen2.5-72B-Instruct-GGUF | quantized | GGUF | — | — | other | second-state | 1 | 54k |
| Gensyn/Qwen2.5-72B-Instruct-bnb-4bit | quantized | -bnb | 74.9B | XL | other | Gensyn | 0 | 49.8k |
| tiny-random/qwen2.5 | finetune | — | 2.46M | S | — | tiny-random | 0 | 40.8k |
| Infermatic/72B-Qwen2.5-Kunou-v1-FP8-Dynamic | quantized | -FP8 | 72.7B | XL | other | Infermatic | 2 | 37.8k |
| Sao10K/72B-Qwen2.5-Kunou-v1 | finetune | — | 72.7B | XL | other | Sao10K | 27 | 35.4k |
| unsloth/Qwen2.5-72B-Instruct | finetune | — | 72.7B | XL | other | unsloth | 5 | 28.7k |
| enfuse/Qwen2.5-72B-Instruct-NVFP4 | quantized | NVFP4 | 42.0B | XL | other | enfuse | 4 | 23.4k |
| mradermacher/Qwen2.5-72B-Instruct-i1-GGUF | quantized | GGUF | — | — | other | mradermacher | 1 | 22k |
| rombodawg/Rombos-LLM-V2.5-Qwen-72b | finetune | — | 72.7B | XL | other | rombodawg | 38 | 19.2k |
| lmstudio-community/Qwen2.5-72B-Instruct-GGUF | quantized | GGUF | — | — | other | lmstudio-community | 5 | 19.4k |
| legraphista/Qwen2.5-72B-Instruct-IMat-GGUF | quantized | GGUF | — | — | other | legraphista | 0 | 19.2k |
| zetasepic/Qwen2.5-72B-Instruct-abliterated | finetune | — | 72.7B | XL | other | zetasepic | 32 | 13.1k |
| Nexusflow/Athene-V2-Agent | finetune | — | 72.7B | XL | other | Nexusflow | 132 | 6k |
| ContextualAI/LMUnit-qwen2.5-72b | finetune | — | 72.7B | XL | — | ContextualAI | 7 | 11.6k |
| mradermacher/Qwen2.5-72B-Instruct-GGUF | quantized | GGUF | — | — | other | mradermacher | 2 | 10.5k |
| CED6688/magnum-v4-72b-AWQ | quantized | AWQ | 73.0B | XL | apache-2.0 | CED6688 | 0 | 9.4k |
| gaianet/Qwen2.5-72B-Instruct-GGUF | quantized | GGUF | — | — | other | gaianet | 1 | 7k |
| miromind-ai/MiroThinker-v1.0-72B | finetune | — | 72.7B | XL | mit | miromind-ai | 129 | 469 |
| axana-labs/Qwen2.5-72B-Instruct-bnb-4bit | quantized | -bnb | 74.9B | XL | other | axana-labs | 0 | 6k |
| AXCXEPT/EZO-Qwen2.5-72B-Instruct | finetune | — | 72.7B | XL | apache-2.0 | AXCXEPT | 12 | 5.3k |
| FreedomIntelligence/HuatuoGPT-o1-72B | finetune | — | 72.7B | XL | apache-2.0 | FreedomIntelligence | 36 | 4k |
| Sakalti/ultiima-72B | merge | — | 72.7B | XL | other | Sakalti | 3 | 5.2k |
| gghfez/Magnum-v1-72b-Qwen2.5 | finetune | — | 72.7B | XL | other | gghfez | 5 | 4.1k |
| Skywork/MindLink-72B-0801 | finetune | — | 72.7B | XL | apache-2.0 | Skywork | 33 | 2k |
| Outlier-Ai/Outlier-150B-V3.2 | adapter | — | — | — | apache-2.0 | Outlier-Ai | 1 | 2.9k |
| allura-org/Qwen2.5-72b-RP-Ink | finetune | — | 72.7B | XL | other | allura-org | 9 | 2.1k |
| tensorblock/Qwen2.5-72B-Instruct-GGUF | quantized | GGUF | — | — | other | tensorblock | 0 | 2.3k |
| amadeusai/Amadeus-Verbo-FI-Qwen2.5-72B-PT-BR-Instruct | finetune | — | 72.7B | XL | apache-2.0 | amadeusai | 2 | 2k |
| Date | Commit | SHA |
|---|---|---|
| 2025-01-12 | Update README.md | 495f393 |
| 2024-09-25 | Update metadata: Add library_name to Transformers (#14) | d3d9511 |
| 2024-09-19 | Upload ./tokenizer_config.json with huggingface_hub | a13fff9 |
| 2024-09-18 | Update LICENSE | 0df185c |
| 2024-09-18 | Update README.md | d167268 |
| 2024-09-18 | Create LICENSE | 347606f |
| 2024-09-18 | update config.json | c970f6d |
| 2024-09-17 | Update tokenizer_config.json | f073433 |
| 2024-09-17 | Update README.md | b07efc4 |
| 2024-09-17 | Update README.md | 2292f49 |
| 2024-09-17 | Update README.md | b1faa85 |
| 2024-09-17 | Create README.md | 64fc727 |
| 2024-09-16 | Update config.json | d9c3167 |
| 2024-09-16 | Upload folder using huggingface_hub | f689aa0 |
| 2024-09-16 | initial commit | 9b34598 |
Hosted endpoints observed for Qwen2.5-72B-Instruct, sorted by input price (USD per million tokens). Each row names its observer and date; prices change daily, so check the source before quoting. Cheapest observed input price: $0.36/M.
| Provider | Model ref | Context | Quant | $/M in | $/M out | tok/s | TTFT | Tools | Status | Observed |
|---|---|---|---|---|---|---|---|---|---|---|
| DeepInfra hf, deepinfra, openrouter | Qwen/Qwen2.5-72B-Instruct | 32K | fp8 | $0.36 | $0.40 | 36.9 | 385 ms | yes | live | 2026-10-26 |
| Featherless hf | Qwen/Qwen2.5-72B-Instruct | — | — | — | — | — | — | — | live | 2026-10-26 |
Ranked by source signals (HN, Lobsters, authority domains, official posts) and recency. Summaries are generated from a stored copy of the article text; every sentence cites a span of that text, and nothing is shown that failed the citation check.
arXiv links found in the model card and its tags. "Peer-reviewed" is shown only with evidence: a published venue in OpenAlex (measured) or the authors' own journal reference on arXiv (reported).
Hub snapshot: 2026-10-26 · HF API: 2026-10-26 07:14 UTC · Providers: 2026-10-26 07:26 UTC · Articles: 2026-10-26 07:27 UTC · Summary: 2026-10-07 15:00 UTC · Last verified 2026-10-26 07:27 UTC
Email alerts are not open yet; the feeds above are the same data. See feeds for the item schema.
